Thought about this too. As for quantum computers, see this post on asymmetric vs. symmetric algorithms. So SHA-512 (and probably even SHA-256) is still safe, and a DMG with AES-256 encryption still offers the same protection as one with AES-128 today.
As for some additional security, I thought about maybe splitting files at a random point before adding it to the IPFS, so e.g. you first encrypt a DMG or archive as strongly as possible, then split it into at least two parts, then add those two parts to the IPFS. Only you, the originator, or a trusted person will know that those two IPFS hashes belong together.
Question is if there is some record kept on when an object was added to the IPFS by which node, and if that record is public. If so, then such a method probably wouldn't work. But if it's an option, you could probably also build it into ipfs, e.g. as
ipfs add --split=2 /path/to/file > will output 2 hashes.
Then you could also extend as
ipfs concat <Hash-1> <Hash-2> <Hash-n> -o /path/to/file